Cirrhosis mortality death rates in Ontario for ages 20 and over declined from a high of 9.3 per 100,000 in 1911 to a low of 5.6 per 100,000 in 1919 (p less than 0.001) and after a 17-year period of relative stability, rose steadily to a high of 19.7 per 100,000 in 1975 (p less than 0.0001) and then declined to 13.3 per 100,000 in 1986 (p less than 0.001). Rates were consistently higher for men than for women and the male to female ratio of the rates increased from a low of 1.3 in 1933 to a high of 2.5 in 1986. The rate of increase in the rates for both men and women, and the rate of decline after the mid 1970s was most noted in the younger ages. Differences in trend could not be related to changes in disease classification, method of recording deaths, changes in diagnostic habits such as introduction of needle liver biopsy or to method of standardizing the rates. There was a positive and significant correlation between per capita alcohol consumption and rates of cirrhosis in Ontario from 1932 to 1975. However, while cirrhosis rates declined markedly from 1976 to 1986, alcohol consumption remained stable from 1976 to 1980 and declined only slightly from 1981 to 1986. A possible explanation for lack of correlation between alcohol consumption and the cirrhosis rates from 1976 to 1986 could be that the balance of force favoured recovery i.e. those people who already had cirrhosis who decreased (or stopped) their consumption of alcohol, did not die. Correlations with lagged alcohol consumption could not explain all the changes in the cirrhosis rates. Although cirrhosis rates consistently increased with increasing age from 35 to 85, our results showed that succeeding generations were developing cirrhosis at successively younger ages after the age of 35. Possible explanations for this cohort effect are increased survival from infectious diseases in infancy and childhood, increase in hepatitis B infection, excessive drinking habits being established at younger ages or a change in the pathogenesis of the disease.